sequence-execution-engine
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The execution engine design allows for data chaining where the output of one step becomes the input of the next. This architecture creates a surface for indirect prompt injection if external, untrusted data is processed without sanitization.
- Ingestion points: The
executeandexecuteParallelmethods inSKILL.mdaccept arbitrary input objects that are passed directly into the step execution functions. - Boundary markers: The library does not provide or specify delimiters, safety prompts, or isolation mechanisms to separate untrusted inputs from the execution context.
- Capability inventory: Steps are defined as arbitrary asynchronous functions; documentation examples include network operations via the
fetchAPI. - Sanitization: The library does not implement built-in validation or sanitization for step outputs before they are passed as inputs to subsequent steps.
Audit Metadata